Judging the efficiency of agricultural machinery operations is the basis for evaluating the utilization rate of agricultural machinery, the driving abilities of operators, and the effectiveness of agricultural machinery management. A range of evaluative factors—including operational efficiency, oil consumption, operation quality, repetitive operation rate, and the proportion of effective operation time—must be considered for a comprehensive evaluation of the quality of a given operation, an analysis of the causes of impact, the improvement of agricultural machinery management and an increase in operational efficiency. In this study, the main factors affecting the evaluation of agricultural machinery operations are extracted, and information about the daily operations of particular items of agricultural machinery is taken as a data source. As regards modeling, a subset of data can be scored manually, and the remaining data is predicted after the training of the relevant model. With a large quantity of data, manual scoring is not only time-consuming and labor-intensive, but also produces sample errors due to subjective factors. However, a small number of samples cannot support an accurate evaluation model, and so in this study a semi-supervised learning method was used to increase the number of training samples and improve the accuracy of the least-squares support vector machine (LSSVM) training model. The experiment used 33,000 deep subsoiling operation data, 500 of which were used as training samples and 500 as test samples. The accuracy rate of the model obtained using 500 training samples was 94.43%, and the accuracy rate achieved with this method with an increased number of training samples was 96.83%. An optimal combination of agricultural machinery and tools is recommended owing to their operational benefits in terms of reduced costs and improved operating capacity.